No abstract

A causal model of the role of general cognitive ability and prior job knowledge in subsequent job-knowledge acquisition and work-sample performance during training was developed. Participants were 3,428 U.S. Air Force officers in pilot training. The measures of ability and prior job knowledge came from the Air Force Officer Qualifying Test. The measures of job knowledge acquired during training were derived from classroom grades. Work-sample measures came from check flight ratings. The causal model showed that ability directly influenced the acquisition of job knowledge. General cognitive ability influenced work samples through job knowledge. Prior job knowledge had almost no influence on subsequent job knowledge but directly influenced the early work sample. Early training job knowledge influenced subsequent job knowledge and work-sample performance. Finally, early work-sample performance strongly influenced subsequent work-sample performance.Several studies have shown the effectiveness of general cognitive ability (g) as a predictor of numerous occupational criteria (e.g., McHenry, Hough, Toquam, Hanson, & Ashworth, 1990;McNemar, 1964;Ree & Earles, 1992). McNemar (1964) and Ree and Earles (1991) demonstrated that g accounted for almost all of the validity of multiple aptitude batteries for predicting training success, whereas McHenry et al. (1990) andTeachout (1994) demonstrated that g was the best predictor of core technical job performance. Hunter (1986) provided a major summary article demonstrating that "general cognitive ability has high validity predicting performance ratings and training success in all jobs" (p. 359).Furthermore, Olea and Ree (1994) extended the aforementioned findings in a large-scale study of about 5,400

K. Pearson (1903) recognized that the correlation coefficient was subject to distortion when a sample was censored or preselected in some way. He proposed 3 univariate correction formulas for better estimates in these circumstances. These have become well known from the work of R. L. Thorndike (1949). D. N. Lawley (1943) proposed a general solution usually called the multivariate correction for range restriction. Both Pearson's and Lawley's corrections are discussed and examples are presented. Of particular interest are the opportunities for the corrected correlations to change sign as a result of the correction. Numerical examples are presented that show that correlations can change signs in the Pearson-Thorndike Case 3 and in Lawley's general solution. Pearson (1903) observed that the correlation coefficient computed in a sample censored in such a way as to reduce variance in one or both of the variables would underestimate the population value. Such censoring that results in a reduction in variance has become known as range restriction. Sometimes the restrictions are so severe that the correlation in the censored sample does not have the same sign as the population correlation. Pearson and Lawley (1943) have provided procedures for correcting these correlations, not all of which allow the sign to change.There is so little published experience in the application of correction procedures that it is not surprising that journal reviewers and members of national oversight committees question results when signs of correlations change as a consequence of application of the procedures. A review of Educational and Psychological Measurement, Journal of Applied Psychology, and Personnel Psychology for the years 1988 through 1992 showed that only 4% of the validity articles were corrected for range restriction. Commonly used statistical packages do not provide procedures to compute either univariate or multivariate corrections. This has contributed to ignorance about correction procedures (Atwater, 1992;Seymour, 1988;Sternberg & Wagner, 1993). Although most researchers have some familiarity with the corrections, knowledge about sign changes is largely absent. This lack of knowledge is the motivation behind the present article.Rectification is past due. An example of a correlation chang-

College students (N= 270) assigned to six‐person mock juries read summaries of a murder trial and then evaluated the defendant's guilt both before and after group discussion. The strength of the prosecution's case was manipulated, as was the inclusion of extra wiretapping evidence that favored the prosecution or the defense and was ruled admissible or inadmissible by the judge. Whether it favored the prosecution or the defense, inadmissible evidence directly biased subjects' reactions toward the defendant and indirectly biased their behaviors during the group discussions. None of these effects varied with the strength of the prosecution's case. The results also showed that the direct effects of inadmissible evidence were at least partially mediated by its indirect effects, suggesting that the process of deliberation can potentially help jurors to control the influence of inadmissible evidence on their decisions.

No abstract